Intervention Effects in Wh-Islands: An Eye-Tracking Study

According to Featural Relativized Minimality (fRM, e.g. Rizzi 2004), the local relation between two elements is disruptedby an intervening one as a function of the syntactic feature overlap between the intervener and the elements it separates:ill-formedness is predicted to be stronger when all syntactic features match (identity; (1),(2)) than when they partiallymatch (inclusion; (3),(4)). According to the Cue-based Memory retrieval model (CM, e.g. Van Dyke and Lewis 2003), re-trieval interference is generated under the same conditions, but both syntactic and semantic features can engender inter-ference. Previous results from acceptability studies on wh-islands returned that configurations with two lexically restrictedwh-elements (2) are more acceptable than those with two bare wh-elements (1), despite both have identical syntacticfeatural specification ([+Q, +N] in (2), [+N] in (1)) (Villata et al. 2013). Although this finding is unexpected under fRM, it isin line with CM, which predicts semantic distinctiveness of restricted wh-elements to engender less retrieval interferencethan bare wh-elements. Although a decrease in acceptability rates has been interpreted as an evidence of a difficulty inthe retrieval of the extractee due to interference, no direct evidence of retrieval difficulty has been provided so far for wh-islands. We present results from an acceptability judgment study in Italian combined with an eye-tracking study on thevery same material with the aim of identifying the locus of the difficulty in the processing of wh-island: if the decrease inacceptability rates is due to retrieval interference, we expect to find evidence of a processing cost at the integration point(the embedded verb or the adjacent spillover region).